This is an interesting proposal. I'm curious to hear any thoughts on how it compares to the protocol in Section 7 of this academic paper: http://crypto.stanford.edu/~xb/fc12/bitcoin.pdf

Both it and CoinSwap appear to enable two party mixing, with unlinkable transactions, providing an anonymity set of all similar transactions happening concurrently. Both depend on what you call hash locking. Both are complex but the complexity is in different spots: in CoinSwap, it requires a number of rounds and a third party Carol. In the paper, it requires a complex setup, including the cut-and-choose.